Causal structure learning from time series: Large regression coefficients may predict causal links better in practice than small p-values

Sebastian Weichwald, Martin E. Jakobsen, Phillip B. Mogensen, Lasse Petersen, Nikolaj Thams, Gherardo Varando
Proceedings of the NeurIPS 2019 Competition and Demonstration Track, PMLR 123:27-36, 2020.

Abstract

In this article, we describe the algorithms for causal structure learning from time series data that won the Causality 4 Climate competition at the Conference on Neural Information Processing Systems 2019 (NeurIPS). We examine how our combination of established ideas achieves competitive performance on semi-realistic and realistic time series data exhibiting common challenges in real-world Earth sciences data. In particular, we discuss a) a rationale for leveraging linear methods to identify causal links in non-linear systems, b) a simulation-backed explanation as to why large regression coefficients may predict causal links better in practice than small p-values and thus why normalising the data may sometimes hinder causal structure learning. For benchmark usage, we detail the algorithms here and provide implementations at {https://github.com/sweichwald/tidybench}. We propose the presented competition-proven methods for baseline benchmark comparisons to guide the development of novel algorithms for structure learning from time series.

Cite this Paper

BibTeX
@InProceedings{pmlr-v123-weichwald20a, title = {Causal structure learning from time series: Large regression coefficients may predict causal links better in practice than small p-values}, author = {Weichwald, Sebastian and Jakobsen, Martin E. and Mogensen, Phillip B. and Petersen, Lasse and Thams, Nikolaj and Varando, Gherardo}, booktitle = {Proceedings of the NeurIPS 2019 Competition and Demonstration Track}, pages = {27--36}, year = {2020}, editor = {Escalante, Hugo Jair and Hadsell, Raia}, volume = {123}, series = {Proceedings of Machine Learning Research}, month = {08--14 Dec}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v123/weichwald20a/weichwald20a.pdf}, url = {https://proceedings.mlr.press/v123/weichwald20a.html}, abstract = {In this article, we describe the algorithms for causal structure learning from time series data that won the Causality 4 Climate competition at the Conference on Neural Information Processing Systems 2019 (NeurIPS). We examine how our combination of established ideas achieves competitive performance on semi-realistic and realistic time series data exhibiting common challenges in real-world Earth sciences data. In particular, we discuss a) a rationale for leveraging linear methods to identify causal links in non-linear systems, b) a simulation-backed explanation as to why large regression coefficients may predict causal links better in practice than small p-values and thus why normalising the data may sometimes hinder causal structure learning. For benchmark usage, we detail the algorithms here and provide implementations at {https://github.com/sweichwald/tidybench}. We propose the presented competition-proven methods for baseline benchmark comparisons to guide the development of novel algorithms for structure learning from time series.} }
Endnote
%0 Conference Paper %T Causal structure learning from time series: Large regression coefficients may predict causal links better in practice than small p-values %A Sebastian Weichwald %A Martin E. Jakobsen %A Phillip B. Mogensen %A Lasse Petersen %A Nikolaj Thams %A Gherardo Varando %B Proceedings of the NeurIPS 2019 Competition and Demonstration Track %C Proceedings of Machine Learning Research %D 2020 %E Hugo Jair Escalante %E Raia Hadsell %F pmlr-v123-weichwald20a %I PMLR %P 27--36 %U https://proceedings.mlr.press/v123/weichwald20a.html %V 123 %X In this article, we describe the algorithms for causal structure learning from time series data that won the Causality 4 Climate competition at the Conference on Neural Information Processing Systems 2019 (NeurIPS). We examine how our combination of established ideas achieves competitive performance on semi-realistic and realistic time series data exhibiting common challenges in real-world Earth sciences data. In particular, we discuss a) a rationale for leveraging linear methods to identify causal links in non-linear systems, b) a simulation-backed explanation as to why large regression coefficients may predict causal links better in practice than small p-values and thus why normalising the data may sometimes hinder causal structure learning. For benchmark usage, we detail the algorithms here and provide implementations at {https://github.com/sweichwald/tidybench}. We propose the presented competition-proven methods for baseline benchmark comparisons to guide the development of novel algorithms for structure learning from time series.
APA
Weichwald, S., Jakobsen, M.E., Mogensen, P.B., Petersen, L., Thams, N. & Varando, G.. (2020). Causal structure learning from time series: Large regression coefficients may predict causal links better in practice than small p-values. Proceedings of the NeurIPS 2019 Competition and Demonstration Track, in Proceedings of Machine Learning Research 123:27-36 Available from https://proceedings.mlr.press/v123/weichwald20a.html.

Related Material